It has been proposed that backward inhibition (BI) is a mechanism which facilitates task-switching by suppressing the previous task. One view is that BI is generated in response to conflict between tasks being experienced during task-performance. Across twelve experiments, this thesis investigated this proposition by addressing two questions: What affects the size/presence of BI? and When is BI triggered?: What affects the size/presence of BI? and When is BI triggered? The findings from Chapter 2 suggest that BI is increased when conflict stemming from shared target features is present, and that the expectation, as well as experience, of conflict might increase BI. Chapter 3 suggests that BI is increased when target features are shared (and that no BI is present otherwise), but contrary to previous findings, BI is not increased when response features are shared. Chapter 4 provided indirect support for the view that BI can be present without between-task conflict (i.e., neither shared targets nor responses), and indicated that in such a context BI (at least at item-level) requires trial-by-trial cuing. Chapter 5 indicates that BI is triggered prior to response execution and after the preparation stage of task processing, therefore indicating that either the target processing stage or the response selection stage of task processing are responsible for triggering BI. Together, the results of the experiments in this thesis indicate that BI can be driven by conflict stemming from target sharing. However, there was no evidence that conflict stemming from response sharing drives BI. In addition, the data suggested that BI might be generated by the expectation of conflict and by task preparation. Therefore, BI might be applied in response to conflict at any stage of task processing and the decision to apply BI might be decided in advance of such conflict.